Mechanisms of

Ice Crystallization
in Ice Cream
Production
K.L.K. Cook and R.W. Hartel

ABSTRACT: The smoothness and perceived quality of an ice cream depends in large part on the small size of ice
crystals in the product. Understanding the mechanisms responsible for producing the disc-shaped crystals found
in ice cream will greatly aid manufacturers in predicting how processing and formulation changes will affect their
product. Because ice cream mix is opaque, it has not yet been possible to observe ice crystallization in ice cream in
situ. Studies to date, therefore, have used analogues or have related observed effects to a hypothesized mechanism.
Still, some elements of the crystallization mechanism are well accepted. Because of the large supercooling at the
freezer wall, ice nucleates there before being swept into the bulk of the freezer. In the bulk, heat and mass transfer
cause some crystals to melt and others to grow. By the time the ice cream reaches the freezer exit, the ice crystals
have become small, rounded discs.

Ice cream is a complex food. . .

- H.D. Goff

Introduction
The delights of ice cream have been known in the western
world since Marco Polo returned from Far East Asia in the 13th
century bringing with him recipes for water ices. Over time, these
water ices evolved into the popular frozen desserts of today. Modern ice cream is a partially-frozen mixture of milk, cream, sugar,
stabilizers, and emulsifiers. Its apparent simplicity belies the complicated structure inside Americas favorite frozen treat. Ice cream
is equally an emulsion, a dispersion, and a foam. The dispersion
and emulsion consist primarily of a freeze-concentrated aqueous serum phase containing sugar and the dry matter contents
surrounding dispersed ice crystals and fat globules (Marshall and
others 2003). Ice crystals range in size from about 1 to over
150 m in diameter, with an average size of about 35 m (Berger
and others 1972; Caldwell and others 1992; Donhowe and Hartel 1996a; Hagiwara and Hartel 1996; Hartel 1996; Koxholt and
others 2000; Marshall and others 2003; Sofjan and Hartel 2004;
Inoue and others 2008; Kusumaatmaja 2009). Fat globules are 2
m or less in diameter (Marshall and others 2003). The foam is
formed by pockets of air (about 20 to 50 m diameter) dispersed
throughout the emulsion and is supported by partially-coalesced
fat globules (Goff 2002; Marshall and others 2003).
The best ice creams on the market have a smooth and creamy
texture. That creamy texture, primarily associated with a high fat
content, is also determined, in part, by the average size of the
ice crystals. Larger crystals (greater than 50 m) impart a grainy
MS 20090967 Submitted 9/30/2009, Accepted 10/25/2009 . Authors are with
Univ. of Wisconsin Madison, 1605 Linden Dr., Madison, WI 53706, U.S.A.
Direct inquiries to author Hartel (E-mail: rwhartel@wisc.edu).


c

R
2010 Institute of Food Technologists

texture to the ice cream, whereas smaller crystals (around 10 to

20 m) give ice cream the desired creaminess (Marshall and others 2003; Eisner and others 2005). With todays consumers becoming more and more health-conscious in their food choices,
low-fat ice cream is becoming increasingly popular. The challenge with low-fat ice cream, however, is achieving the same
creamy texture of the full-fat version. With less fat, ice cream commonly contains more water and, thus, larger ice crystals (Hartel
1996).
Ice crystal size is also important to ice cream shelf life. As
the ice cream sits in storage, the ice crystals continually grow
by recrystallization (Donhowe and Hartel 1996b; Hartel 1998a).
The temperature fluctuations seen in a frost-free home freezer
serve to accelerate this process.
The actual mechanism of ice crystallization in the scrapedsurface heat exchanger remains the subject of research due to
the technical difficulties in watching ice form through opaque
ice cream mix and inside a scraped-surface freezer. Some basics
of the mechanism, however, are generally well accepted. At the
wall of the ice cream freezer, and only at the wall, the temperature
is cold enough to form new crystals. A layer of ice forms on the
wall with properties depending on mix composition and freezer
operation. The dasher blade then comes around and scrapes the
ice off the wall, dispersing the frozen layer into the bulk of the
freezer. According to rates of heat and mass transfer in the center
of the freezer, some crystals melt while other crystals grow in a
complex process called ripening. The final result at the freezer
exit is a distribution of small, disc-shaped crystals.
Understanding how ice forms during the freezing of ice cream
will greatly aid manufacturers in predicting the effects of processing and formulation changes. The purpose of this review is
to discuss research on how ice crystals form in a scraped-surface
freezer and the parameters that affect their growth. Brief descriptions of general crystallization and ice cream manufacture are
presented first to provide background.

Vol. 9, 2010COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETY

213

CRFSFS: Comprehensive Reviews in Food Science and Food Safety

Crystallization
Crystallization is the organization of molecules into a solid
phase within a fluid. This can occur in several different ways.
Vapor can crystallize directly from the gas to solid phase, with
the prime example being snow. A solute can be crystallized out
of solution, such as in lactose crystallization in dairy products.
Alternatively, the solvent can crystallize, as happens when ice
forms in ice cream. For a mixture of liquids, lowering the temperature can cause crystallization based on the phase diagram
of the mixture, such as in metallurgy (McCabe and others 2005).
Since ice crystallization out of a melt is the type of crystallization
of concern in ice cream manufacture, it will be the focus of the
rest of the discussion on crystallization, though the same general
principles apply to other types of crystallization.
The structure of ice determines how the water molecules will be
organized after crystallization. Typical ice is categorized as ice Ih,
which is hexagonal and prismatic in form and occurs below 0 C
at atmospheric pressure (Fennema 1973). While there are 9 other
forms of ice, most exist at pressures greater than 200 MPa. Ice has
a hexagonal structure, described by 3 a-axes in one plane and
a c-axis normal to that plane (Schwartzberg 1990). Experiments
involving freezing sugar solutions on a cold surface have shown
the a-axes to grow faster than the c-axis (Schwartzberg 1990). The
same is true for snow flakes, which are well-known to exhibit a
hexagonal structure (Libbrecht 2005).
Before crystallization can begin, the temperature of the melt
must decrease below the freezing point. When the melt is a
solution, the solute lowers that freezing point to an extent based
on the nature of the solute. Freezing point depression, being a
colligative property, is based on the moles of solute present in
solution, not on the mass. Therefore, a gram of glucose (with
about double the number of moles of sucrose) would lower the
freezing point more than a gram of sucrose. Proteins, being very
large comparatively, have no measurable effect on freezing point
(Marshall and others 2003).
The size of a crystal can also affect the freezing point. While
ice Ih has a nominal freezing point of 0 C, smaller ice crystals are
known to have a slightly depressed freezing point compared to
larger crystals. The same is also true for a sharp point on a crystal
surface compared to a flat surface. This phenomenon is due to
what is called the Kelvin or radius of curvature effect. Because
the small crystals have more surface free energy per volume than
do the larger crystals, their freezing point is depressed according
to Eq. 1 (Schwartzberg 1990; Hartel 2001).
T  =

2 T
s Hfus r

(1)

Here, r is the crystal radius, T  is the freezing point depression,

T is the equilibrium absolute temperature for a crystal of infinite radius, is the interfacial tension (1.5 to 3.2 102 J/m2 )
(Hobbs 1974), s is the density of the crystal (917 kg/m3 for ice)
(Fennema 1973; Schwartzberg 1990), and Hfus is the latent heat
of crystallization (Schwartzberg 1990).
Crystallization occurs in 4 steps: supercooling, nucleation,
growth, and recrystallization. Supercooling occurs when the temperature of the liquid drops below its freezing point. In general,
water molecules are constantly coming together to form clusters
held together by hydrogen bonds. These clusters are continually
breaking up again due to molecular vibrations, although lower
temperatures enhance cluster formation (Schwartzberg 1990;
McCabe and others 2005). At a certain level of supercooling,
the ice-like clusters gain enough molecules of water that they
have an equal chance of surviving or not as crystal nuclei. Thus,
214

there is a critical radius above which the nuclei will be likely to

survive (Schwartzberg 1990).
In a scraped-surface freezer, crystallization conditions can be
divided into 2 zones. Close to the wall, the fluid is very cold, so
nucleation occurs in that zone only. Far from the wall, in the bulk
of the fluid, temperatures are closer to the freezing point, and
growth and recrystallization predominate (Cebula and Russell
1998).
Nucleation

Nucleation is the birth of a crystal (Hartel 2001). In a scrapedsurface freezer, this only occurs at the wall of the freezer where
the temperature gradient is high enough to form new crystals.
The rate of nucleation is important because it affects the number
of crystals in a product, which in turn affects how large each
crystal can become. For ice cream, once it exits the freezer, the
maximum number of ice crystals is set, and those crystals continue to grow until the amount of ice is in equilibrium with the
temperature (Hartel 1996).
The nucleation process starts with the initial cluster of ice as a
loose aggregation of molecules. As the cluster loses energy, the
molecules become more ordered and a lattice structure begins
to form and the cluster becomes an embryo. When the embryo
grows large enough to be in thermodynamic equilibrium with the
solution and when the embryo overcomes the energy barrier to
nucleation, it is a nucleusthe smallest collection of molecules
in a crystal lattice that will not redissolve. When the nucleus
grows it becomes a crystal (McCabe and others 2005).
Nucleation can occur in several ways. Primary nucleation describes the spontaneous formation of a nucleus from solution,
whereas secondary nucleation describes the formation of a nucleus from a preexisting crystal or crystal fragment. There are 2
types of primary nucleation: homogeneous and heterogeneous.
In homogeneous nucleation, the nucleus is formed in the bulk
of the fluid by molecular accretion (McCabe and others 2005).
In heterogeneous nucleation, on the other hand, the nucleus
forms on a foreign particle or object, which could be anything
from a dust mote to the wall of an ice cream freezer (Hartel
2001). Because heterogeneous nucleation occurs more readily
and with less supercooling than does homogeneous nucleation,
it is much more common in practical applications (Schwartzberg
1990). Secondary nucleation most commonly occurs in industry as contact nucleation, which is when a crystal collides with
another crystal, the wall, or an agitator (McCabe and others
2005).
Several factors affect nucleation. The most important factor is
the driving forcethe temperature difference between the solution and its freezing point. A greater driving force increases nucleation. However, when the temperature is too low, decreased
molecular mobility slows nucleation. Below the glass transition
temperature, T g , no nucleation occurs. Agitation can enhance nucleation in a labile (ready-to-nucleate) system. Cooling rate influences nucleation as well. For example, if no nucleation is desired,
the solution can be quickly cooled to a temperature below the
glass transition before the onset of crystallization (Hartel 2001).
In foods, the formulation and ingredients can impact nucleation;
for instance, sucrose may slow ice nucleation (Schwartzberg
1990).
Growth

After ice nucleates at the wall of an ice cream freezer it must

grow to become the disc-shaped crystals found in the exit stream.
For crystal growth to occur, water must diffuse to the growing crystal surface, solutes must diffuse away from the surface,

COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETYVol. 9, 2010

Mechanisms of ice crystallization in ice cream . . .

water molecules must incorporate themselves into the crystal lattice structure, and latent heat must be removed (Fennema 1973;
Schwartzberg 1990; Hartel 2001). While the internal structure of
the ice is generally uniform, growth conditions have a profound
effect on the external shape and sizethe morphologyof the
crystal. Generally speaking, slow freezingslow transport of latent heatcauses fewer nuclei to form per a given volume and
time, which allows more space for each crystal to grow in and results in larger crystals. Quick freezing results in a faster nucleation
rate and a larger number of small crystals (Schwartzberg 1990;
Hartel 2001). Normally, dendritic (tree-like), needle-shaped, and
fibrous crystals form with quick freezing, whereas slow freezing allows uniform growth into smoother shapes (Schwartzberg
1990). Dissolved solutes also impact crystal morphology by adsorbing to the crystal surfaces, which interferes with the lattice
incorporation step, and by decreasing molecular mobility (increasing viscosity) (Fennema 1973).
Dendritic and branched morphologies are common in crystals
made with high supercoolingsuch as at the wall of an ice cream
freezer. Before dendritic structure evolves, a growing ice crystal
is approximately circular (Teraoka and others 2002). As one section of the crystal begins to protrude, that tip is surrounded by a
higher water concentration, so crystallization is easier at the tip.
Dendrites evolve into a branched structure with primary branches
developing secondary branches and on to higher-order branches.
This positive-feedback loop is called the MullinsSekerka instability (Libbrecht 2005).
Dendrite morphology is dependent on the behavior at the crystal tips. This behavior is characterized by the dendrite growth velocity and the tip radius (the radius of a circle fitting the curvature
at the tip), which are in turn a function of the supercooling and
the solids concentration (Hindmarsh and others 2005). A perfect dendrite will have a hexagonal shape, but at higher growth
velocities (higher supercooling) an increased number of dislocations can cause crystals that have extra branches parallel to the 6
main branches (Teraoka and others 2002). The spacing between
dendrites has been found to be proportional to the freezing rate
to the 12 power (Schwartzberg 1990), which means that dendrites formed at higher supercooling have closer (and thinner)
branches. As well, higher solute concentration produces thinner
branches (Teraoka and others 2002). Between 2 branches, there
will be stagnant regions with high solute concentration and low
freezing point. As a dendrite ripens these branches may become
thicker and more widely spaced (Huang and Glicksman 1981a).
The branches in dendrites are observed to be at 60 angles, indicating that the a-axes of the ice grow faster than the c-axis
(Schwartzberg 1990). This has been observed in hexagonal ice
crystals growing laterally out of a sucrose solution before growing
vertically (Si 2000).
While an instability easily results in branch formation, the
freezing point at the tip is also lower, so dendritic growth is only
favored with high supercooling (Schwartzberg 1990). The heat
being removed from the crystal tip is conducted away mainly
through the ice because ice is about four times as conductive
as water (McCabe and others 2005). At steady state, the dendritic stability criterion states that the tips growth velocity times
the square of the tip radius is constant (Huang and Glicksman
1981b).
A negative thermal gradient at the freezing front can result in
irregular or dendritic crystal growth. In this case, temperature decreases from the interface both into the liquid and into the solid
(temperature is highest at the interface). This may occur if the rate
of heat removal is greater than the rate of crystallization (Fennema
1973). The heat of crystallization can then be conducted away
through both the crystal and the liquid. Also, because the temper-

Figure 1 --- Conditions for constitutional supercooling

(Fennema 1973).

ature is lower farther away from the interface, there is a stronger

driving force for a protuberance from the crystal to grow into
the solution. Secondary protuberances will cause the branches
typical of dendritic crystals (Fennema 1973; Zheng 2006).
A positive temperature gradient on the other hand, where temperature continues to increase from the solid to the interface and
into the liquid, discourages dendritic growth. The heat of crystallization is conducted away through the crystal only, so any
protruding crystal will not only be in a warmer location but will
also need to conduct its heat of crystallization farther back to the
main crystal (Fennema 1973; Zheng 2006).
At the crystallization front, the solution will be more concentrated due to counter-diffusion of solutes. As a result of the higher
concentration, the freezing point temperature is reduced at the
surface, with a gradient that levels off to the bulk freezing point
farther away from the crystal surface. If at the same time the
temperature in the solution increases from the freezing point at
the crystal surface to the bulk temperature more slowly than the
freezing point increases, the result will be constitutional supercooling, as shown in Figure 1 (Fennema 1973; Zheng 2006).
Constitutional supercooling differs from normal supercooling in
that it is due to a solute concentration gradient. Within the region of constitutional supercooling, protrusions may grow into a
dendritic form, as with a negative temperature gradient (Fennema
1973; Zheng 2006). If 2 crystals are close together, the higher solute concentrations at each interface may overlap, flattening and
lowering the freezing point curve such that the constitutional
supercooling region disappears and melting occurs. Fast solidification and low solute diffusivity, as found in ice cream freezing,
favor constitutional supercooling (Zheng 2006).
Recrystallization

Recrystallization, also called ripening or coarsening, is the

final stage of crystallization describing a change in the shape
and size of the individual crystals while keeping the total crystal
mass constant. It can occur both during a crystallization process,
such as during initial ice cream freezing, and after the process, such as during storage of ice cream. Recrystallization, like
growth, is a function of heat- and mass-transfer, as well as the
Kelvin effect. It occurs because a system tends towards an equilibrium with minimal free energy (Schwartzberg 1990). Also, because of the Kelvin effect, small crystals and sharp points on a
crystal (greater freezing point depression) melt faster than larger,
rounder crystals at a given temperature (Hartel 1998b).
The 3 main types of recrystallization are isomass, migratory, and accretive recrystallization. Isomass recrystallization
is the change in shape of a crystal without change in mass.

Vol. 9, 2010COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETY

215

CRFSFS: Comprehensive Reviews in Food Science and Food Safety

Migratory recrystallization, or Ostwald ripening, describes large
crystals growing at the expense of small crystals. Accretion is
2 crystals joining together to form 1 crystal (Fennema 1973;
Donhowe and Hartel 1996b).
Recrystallization is often modeled according to Eq. 2, where
r is the average radius of a spherical particle at time t, r0 is the
initial average radius, and and n depend on the mechanism
(Hartel 1998b; Flemings 2005).
r n = r0n + t/

(2)

For long times, recrystallization can be modeled according to

Eq. 3, where R is the rate of recrystallization (Hartel 1998b).
r = r0 + Rt 1/n

(3)

The value of n can reveal the mechanism of recrystallization.

For n = 3, ripening is said to be limited by diffusion of the solute
from the bulk in a stagnant system (Lifshitz and Slyozov 1961;
Wagner 1961). A value of 2 for n indicates either that ripening is
controlled by the lattice incorporation (Jain and Hughes 1978) or
surface reaction step (Mullin 1993) or that convective diffusion
limits growth (Kahlweit 1975). When n = 4, grain boundary diffusion has been found to be limiting; that is, the grain boundaries
serve as the active sites for diffusion from the solution (Jain and
Hughes 1978). The rate of recrystallization itself is a function of
the system properties, such as composition and temperature. For
hardened ice cream in storage, recrystallization follows Eq. 3 with
n equal to 3 and the rate of recrystallization has an Arrhenius-type
dependence on temperature (Donhowe and Hartel 1996a).
Recrystallization is not only a function of time, however. The
crystal size distribution can also play a role. For a given distribution at a given temperature, there may be a few crystals whose
radii put them in perfect equilibrium with the bulk temperature.
Other crystals will be either smaller or larger than the equilibrium radius, and the small crystals will tend to melt due to their
lower freezing point, and the large crystals will grow (Smith and
Schwartzberg 1985). Melting the small crystals takes energy out
of the solution, and this heat removal, in turn, provides a heat
sink for the latent heat of crystallization given off from the growing larger crystals. While growth is governed by heat and mass
transfer and lattice incorporation kinetics, crystal dissolution or
melting is mainly governed by heat and mass transfer. Manipulation of the size distribution can affect how fast the system
ripens (Smith and Schwartzberg 1985; Sutton and others 1994).
For example, a size distribution where every crystal was within
a couple microns of the equilibrium radius would ripen very
slowly, whereas a very wide distribution would ripen faster because the crystals at each extreme will have a larger driving force
for recrystallization.
When freezing ice cream, dendrites may form initially
(Schwartzberg and Liu 1990); however, by the time the ice cream
exits the freezer, the crystalswhatever their initial formhave
become disc-shaped. How dendrites recrystallize has been wellstudied in the field of metallurgy for solidification of metal alloys.
As a dendrite coarsens, branches separate from their trunks, the
thinnest branches melt away, the trunks increase in width, and all
the points become more rounded. Dendrites are known to ripen
in 3 ways (Flemings 2005). First, large dendrite arms may grow
at the expense of small dendrite arms. Second, the arms may
coalesce as the space between the dendrite arms fills in and
the dendrite starts becoming rounder. Third, the number of dendrites, but not the mass, may multiply as the crystal breaks apart.
Fluid motion in the early stages of solidification causes dendrites
to break up, although increasing fluid velocity does not always
raise the breakup rate (Flemings 2005; Mullis 2006). After the
216

arms separate, they become disc-shaped (Flemings 2005). One

possible mechanism for dendrite breakup is that the dendrites
separate at their branch points, which is where the radius of curvature is high and the crystal is prone to melting (Flemings 2005;
Mullis 2006). In quiescent systems, dendrites may exist intact for
a long time. However, any perturbation may cause the dendrite to
break apart by partial melting and arm detachment (Mullis 2006).
Often, many dendrite arms will detach at once, possibly due to
a small energy input, such as slight agitation, driving the system
farther from equilibrium and increasing the melting rate (Mullis
2006). Convection then takes away any superheat into the bulk
along with the dendrite fragments (Flemings 2005).
When quantitatively characterizing dendrites, the distance between secondary dendrite arms (branch spacing) and the dendrite
tip radius are common measurements of fineness. A finer, thinner
dendrite will have arms spaced close together and a smaller tip
radius. As the dendrite coarsens, the arms will become fewer and
farther apart and the tip radius will increase. Dendrite coarsening
has also been modeled with Eq. 3 with n equal to 3 and r as the
tip radius (Glicksman and Voorhees 1984).

Manufacture of Ice Cream

To produce ice cream in the typical commercial process, the
ingredients are first mixed together, then pasteurized, homogenized, cooled, and aged for at least 4 h at a temperature of about
4 C. Flavoring is then added before the mix is pumped into the
scraped-surface freezer (SSF), where about 50% of the water is
frozen and air is incorporated into the product. Upon exiting
the SSF, inclusions and variegates may be added, and the soft
product, at about 5 to 6 C, is filled into its retail container.
The ice cream is sent to harden until the core reaches a specified temperature, usually about 18 C. As the temperature drops
and more ice crystallizes, the remaining water contains more and
more sugar, which depresses the freezing point so that in the final,
hardened product, about 75% of the water is frozen. The finished
product is stored between 18 and 30 C (Ben-Yoseph and
Hartel 1998), depending on the plant, distributed, and sold (Hartel 1996; Marshall and others 2003). In the supermarket, temperatures in certain types of freezer cabinets (particularly open-faced
cabinets) can reach 9 C, and in frost-free consumer freezers,
temperatures can vary quite widely during the frost-free cycle
(Ben-Yoseph and Hartel 1998).
Initial freezing in the SSF is arguably the most important step
in creating ice cream. This is the only step in which ice crystals
are formed. During hardening, the crystals formed in the SSF
grow to accommodate the increased crystal mass. Assuming an
equilibrium ice content in the draw (exit) stream of the SSF, the
more ice crystals there are at draw, the more ice crystals will be
preserved during hardening, leading to a smaller average crystal
size. Because of this, it is critical to product quality to control and
optimize ice nucleation, growth, and ripening in the SSF. Ideally,
the ice crystals will be kept as small as possible to enhance
creaminess and extend shelf life.
A typical SSF consists of a cylindrical barrel with a refrigerant,
such as vaporizing ammonia or Freon, surrounding it. Inside the
barrel is a rotating shaft, or dasher, with scraper blades attached to
it. Typically, the barrel and dasher assembly are made of stainless
steel (Marshall and others 2003). Ice cream may be frozen either
in a batch or continuous process. A batch process may take on the
order of 10 to 20 min to freeze. A continuous freezer, common
to large operations, has a residence time on the order of 1 to
2 min (Goff and Hartel 2006). The temperature profile inside
a continuous freezer is not uniform. Radially, the temperature
increases from 26 to 28 C at the wall, where the refrigerant
temperature is about 30 C, up to the bulk fluid temperature at

COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETYVol. 9, 2010

Mechanisms of ice crystallization in ice cream . . .

Figure 2 --- Three-dimensional schematic diagram of ice crystallization and fluid flows through a scraped-surface heat
exchanger for ice cream manufacture. The scraper blade removes a slush of ice crystals and concentrated ice cream
mix from the barrel wall in flocs, the dasher breaks apart these ice flocs, and the warm bulk liquid melts the flocs
apart into disc-shaped crystals. Near the inlet, the warm bulk temperatures melt the majority of the ice. As freezing
continues more ice crystals survive in the bulk. Close to the exit, the bulk crystals are larger; the wide size range
causes many small new crystals to melt.

the center (Hartel 1996). Axially, from the inlet to the outlet along
the barrel center, the temperature has been shown to decrease
rapidly in the first 15% of the barrel to between 3 and 7 C
and then decrease linearly to just before the freezer exit, where
the temperature increases by about 1 C (Hartel 1996; Russell
and others 1999).
Initial freezing is a dynamic process: the mix is frozen while
being agitated, which whips in the air, destabilizes the fat, and
scrapes ice into the bulk fluid (Goff 1997; Marshall and others
2003). As illustrated in Figure 2, cold ice cream mix enters the
freezer, with the refrigerant absorbing the heat in the mix until
the supercooling at the wall is great enough to initiate ice nucleation. As ice crystals form at the wall, the dasher blades scrape
the surface layer from the wall about every 0.075 s, assuming 4
blades with a dasher speed of 200 rpm, and propel the ice layer, at
a temperature of 20 to 25 C (Hartel 1996), into the bulk of the
freezing ice cream mix. This is pictured schematically in Figure 2,
with a larger, cross-sectional view shown in Figure 3. In the bulk,
the warmer temperatures cause the ice crystals to ripen into disclike shapes. The details on how the ice crystals actually form
and ripen is still the subject of research, although several theories
have been proposed (see next section). When the ice cream exits the freezer, it is usually around 5 to 6 C, and close to
half of the water is frozen. From a survey of recent studies, the
average ice crystal size exiting an ice cream freezer is 20 to 30
m (Donhowe and Hartel 1996a; Hartel 1996; Russell and others 1999; Chang 2000; Windhab and Wildmoser 2002; Marshall
and others 2003; Sofjan and Hartel 2004; Drewett and Hartel
2007).
Based on the axial temperature profile in a SSF, ice cream
freezing generally follows the solid-liquid equilibrium curve on
an ice cream mix phase diagram. Figure 4 shows an approximate phase diagram for ice cream mix. The actual freezing point
for an ice cream mix can be calculated from published equations (Bradley 1984; Marshall and others 2003) and is based
on the type and content of sweeteners and other solids in the
mix. As freezing begins, the ice crystals formed are purely frozen
water (Omran and King 1974), which leaves the solutes dissolved
in less water than before. As freezing continues and the serum
phase is further concentrated, the freezing point decreases until
the target temperature has been achieved. If the ice cream is fur-

Figure 3 --- Two-dimensional view looking down the length

of the dasher showing a larger view of ice crystallization
near the freezer exit. At this stage, the ice crystals ripening in the bulk are mostly larger than the ice discs formed
from the material scraped off the freezer wall, likely resulting in the large crystals growing at the expense of
the smaller crystals. For clarity, the size of the crystals
inside the slush and ice flocs has been exaggerated.

ther cooled to below the glass transition temperature, Tg , it has

reached the maximal freeze concentration and becomes a glass.
However, because ice cream is stored at temperatures above the
glass transition temperature, normal storage is between 10 and
20 C, the ice can continue to change and recrystallize, mainly
by isomass rounding and accretion (Donhowe and Hartel 1996b;
Hartel 1996).
Although no new ice crystals nucleate after the freezer (heat
transfer is not fast enough), the existing ice crystals continue to
grow as the product cools from the draw temperature to 18 C.
Typically during hardening, ice crystals grow by about 30% to
40% (Marshall and others 2003) to an average size of about 25
to 45 m (Berger and others 1972; Caldwell and others 1992;
Donhowe and Hartel 1996a; Hagiwara and Hartel 1996; Koxholt
and others 2000; Sofjan and Hartel 2004; Inoue and others 2008;
Kusumaatmaja 2009). Because higher temperatures accelerate
recrystallization, quick hardening limits the growth of the ice
crystals (Hartel 1996; Marshall and others 2003).

Vol. 9, 2010COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETY

217

CRFSFS: Comprehensive Reviews in Food Science and Food Safety

Figure 4 --- Sample phase diagram for ice cream mix containing 10% milk solids nonfat, 16% sucrose, and 38%
total solids; Tg is the glass transition temperature at
 is the end of the freezmaximal freeze concentration, Tm
ing point curve, and Cg is the maximal freeze concentration (Hartel 1996; Marshall and others 2003; Schawe
2006). The dashed lines indicate that there is a degree
of uncertainty when predicting the glass transition curve
and the freezing curve for ice cream mix with high solids.

Freezing Mechanisms
The specific mechanisms of ice crystal formation in a scrapedsurface heat exchanger during ice cream manufacture remain
somewhat elusive. No one has yet been able to observe ice crystals forming while making ice cream, mainly because ice cream
mix is opaque and the freezer barrel is enclosed. Various studies, therefore, have resorted to either using sucrose solutions as
model systems for ice cream (Omran and King 1974; Stocking
and King 1976; Schwartzberg and Liu 1990; Sodawala and Garside 1997; Jacobsen 1999; Si 2000; Lian and others 2006; Zheng
2006) or to examining cause-and-effect relationships when
making ice cream in an SSF (Cebula and Russell 1998; Russell
and others 1999; Drewett and Hartel 2007).
Crystal morphology is of fundamental importance to describing
how ice forms in an SSF. If the crystals are initially disc-shaped,
they will ripen differently than if the ice leaves the freezer wall
in dendritic form. Numerous studies have attempted to define
the initial ice crystal shape formed in an SSF. In freezing sugar
solutions, disc-shaped crystal morphologies were observed at
very low supercooling (under 100 mK). At higher supercoolings,
such as in an SSF, however, dendritic or needle-like morphologies occurred (Stocking and King 1976). In quiescently-frozen
sugar solutions, as well, dendritic growth is well-known to occur
(Macklin and Ryan 1966; Omran and King 1974). At constant
supercooling, increasing the sugar concentration delays nucleation and causes crystal size to decrease (Omran and King 1974;
Schwartzberg 1990). For dendrites, increasing sugar content also
causes thinner branches (Schwartzberg 1990).
Based on the principles of dendrite formation, Schwartzberg
proposed the lawnmower model. Schwartzbergs experiments
involved observation of sucrose solution quiescently-frozen on a
chilled surface (Schwartzberg and Liu 1990). Because the experiments used quiescent freezing, dendritic growth was observed.
However, he suggested that due to the high supercooling at the
218

wall in an ice cream freezer, dendritic ice growth is likely there,

as well. In this model, dendrites grow normal to the freezing
surface, and when a scraper blade comes around, it cuts off the
budding dendrites at their base on the wall. Fluid motion transports the dendrites into the bulk of the freezer, where they ripen
into disc-shaped ice crystals (Schwartzberg 1990).
The heat input during scraping may also affect ice formation in
an SSF. Schwartzberg (1990) calculated that the dasher inputs a
large amount of frictional heat into the freezeras much as 50%
of the total heat removedwhich will cause some crystals to
melt and will slow the freezing process. Russell and others (1999)
measured the actual energy input from the dasher, which showed
that about 15% to 40% of the heat removed was frictional heat,
depending on the dasher speed and draw temperature. Clearly the
dasher inputs a high amount of heat into the freezer, which likely
has a significant, yet largely unknown, effect on ice formation in
an SSF.
Cebula and Russell (1998) considered ice cream freezing in
an SSF to occur in 2 distinct zones: a wall zone and a bulk
zone. In the wall zone, they found, by spreading mix over a
cold plate and then examining a cross-section with a scanning
electron microscope, that globular crystals formed in the vicinity
of the surface and that columnar crystals formed farther away. The
columnar crystals did not attach themselves to the wall, however,
but grew out from small nuclei near the cold wall. When the
surface was scraped, it was observed that remaining ice debris
served as secondary nuclei for new crystal growth. The new ice
grew in roughly hexagonal shapes, which may have had time to
grow into a unified layer depending on the length of time between
scrapes. Investigating the bulk zone, Cebula and Russell (1998)
varied the dasher speed, which varied the specific net energy
dissipated by the freezer, while keeping residence time constant,
and found that increasing the total energy input greatly increased
the average crystal size. This likely occurred because the extra
energy in the system would have melted many of the smallest
(and least stable) crystals, leaving fewer and larger crystals in the
product. They concluded that, in the wall zone, the mechanism
of ice formation is probably somewhere between heterogeneous
nucleation at the wall and nucleation from ice debris and that, in
the bulk zone, the total energy input into the system is critical to
the final crystal size distribution.
A subsequent study by Russell and others (1999) showed that
product residence time in the freezer had a more important
impact on final crystal size distribution than did initial nucleation rate. Ice creams made by varying the wall temperature
while keeping the residence time constant gave very similar crystal size distributions, whereas ice creams made with a constant
exit temperature (constant heat removal) or a constant coolant
temperature with varied residence time gave different distributions. A shorter residence time gave a smaller average crystal
size (Russell and others 1999). The fact that residence time had a
more significant impact on the final average crystal size than did
the wall temperature indicates that ripening is more significant
than nucleation. A similar study by Drewett and Hartel (2007)
also concluded that residence time had the greatest impact on
final crystal size distribution, followed by draw temperature and
dasher speed.
Sodawala and Garside (1997) observed freezing of a 10% sucrose solution with a videomicroscope in a flowcell simulating
a scraped-surface heat exchanger. The flowcell contained a flat
stainless steel surface cooled underneath by a coolant and a
sharp-edged blade that scraped the surface at intervals while
sucrose solution flowed over the metal surface. After a scrape,
islands of ice grew parallel to the freezing surface before merging
and continuing to grow normal to the surface. At low scraping
frequencies, flocs of ice were scraped off the cold surface into the

COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETYVol. 9, 2010

Mechanisms of ice crystallization in ice cream . . .

bulk of the solution. Higher scraping frequencies generated more
heat, which reduced both nucleation and growth rates; they also
resulted in smaller flocs being scraped off.
Studying ice crystallization from a sucrose solution in a slush
freezer, Jacobsen (1999) found that oblong ice crystal sheets were
initially scraped off the heat-exchange surface by the blade. Agitation within the warmer bulk fluid then broke apart these crystal
sheets, which ripened into rounded ice crystals within 10 min.
After 16 h in the slush freezer, pellet-shaped crystals had formed,
as opposed to the typical flat disc-shaped crystals found in hardened ice cream.
Ice crystallization and transformation were observed under
many different experimental conditions in an effort to elucidate how ice forms in an SSF (Si 2000). Sucrose solution was
frozen on a microscope slide placed over a small glass flowcell containing refrigerant inside a refrigerated enclosure. A hightemperature driving force between the refrigerant and the slide
caused dendritic crystals to grow. Without agitation at a constant temperature, the dendrites did not become discs. In another
experiment, 40% sucrose solution was frozen without stirring
in a chilled, jacketed glass beaker. It was observed that at a low
wall temperature, the dendrites grew closer together and that near
25 C some even looked glassy. Subsequent warming from 20
to 10 C caused dendrites to subsequently melt into many, discshaped crystals. A subsequent experiment observed ice crystallization in a scraped-surface heat exchanger with a sucrose and
gelatin solution. Large (100 to 450 m) hexagonal crystals were
seen to form initially and remain for about 20 s before changing
into a dendritic form. The hexagonal crystals grew quickly for the
first 5 s, after which crystal growth was slower. The scraper blades
then broke the dendrites into shreds (Si 2000). For a short time
the shreds and dendrites could coexist. Then, the shreds partially
melted in the bulk of the freezer and eventually rounded into
discs (Si 2000).
More recently, using a flowcell equipped with a scraper blade
and a chilled surface, Zheng (2006) studied heat transfer during
ice crystallization from a sucrose solution. Zheng observed that
many nuclei rapidly grew from ice debris remaining after each
scrape and continued growing along the cooling surface before
merging and growing vertically, as also seen by Sodawala and
Garside (1997). It was then concluded that, for crystallization,
the most significant effect of the scraper is to cause secondary
nucleation at the surface. By monitoring the heat flux through
the cold surface, Zheng (2006) showed that crystallization at the
surface greatly enhanced heat transfer because the heat could
travel directly from the crystal, through the wall, and to the refrigerant, without having to travel through the liquid. The frozen
layer at the cold surface was determined to be effectively a slush
composed of both ice and concentrated sucrose solution. This
was concluded when thermal conductivity measurements of the
frozen layer yielded a value close to the average thermal conductivity, weighted by volume fraction, of ice and the concentrated
sucrose solution. A slush requires less power to scrape than a
layer of solid ice, which suggests reduced scraping power might
be possible (Zheng 2006).
Lian and others (2006) modeled ice crystallization in an SSF for
sucrose solution. They used computational fluid dynamics coupled with crystal population balance to predict the ice crystal size
distribution, assuming only nucleation and growth and no aggregation, breakage, or back-mixing. The model results matched the
shape of the ice crystal distribution curve, but overestimated
the mean crystal size (Lian and others 2006). It is possible that
the overestimation is a result of not accounting for effects that
would shrink crystal size, such as breakage and melting.
Pulling together the findings from these diverse studies, a basic
mechanism for ice crystal formation in a scraped-surface freezer

may be hypothesized. First, an ice layer forms on the wall, and

the scraper blade removes the ice layer into the bulk. Exactly
how this layer forms is still debatable. Depending on how well
the scraper blade removes ice from the wall, there may be a little
residual ice left behind, if only in microscopic crevices, to seed
the next round of crystal nucleation, or there may a relatively
thick layer of ice left behind that would simply keep growing.
Another factor is what happens when, after a scrape, the icecontaining bulk fluid comes in contact with the wall. Perhaps the
cold wall-zone temperatures cause the existing ice to grow at an
accelerated rate, possibly in a dendritic fashion. Based on video
images from the research of both Zheng (2006) and Sodawala and
Garside (1997), it is likely that the ice growing at the wall grows
parallel to the freezing surface. Given enough time for the ice
growths to converge, the ice would then begin to grow normal
to the surface. Regardless of exactly how it forms, the ice layer
scraped off the wall is probably a slush containing dendrites
and entrapped, concentrated ice cream mix. Once the scraper
blade disperses the slush layer at the wall into the warmer bulk
medium, the result would be the crystals melting apart into many
disc-shaped crystals. Toward the freezer entrance, much, if not
all, of the ice dispersed into the warmer bulk fluid is likely to melt
and bring down the bulk temperature. Farther along the length of
the freezer, more ice would survive. By the freezer exit, the bulk
ice crystals would be much larger than the new crystals being
transported from the wall. This wider crystal size distribution
would accelerate ripening and encourage the large crystals to
grow at the expense of the small crystals. Energy input from the
dasher, due to friction at the wall and viscous dissipation, also
can contribute significantly to crystal melting and ripening.

Factors Affecting Crystallization

The course of crystallization is determined by many factors that
interact with and depend upon one another. Ice cream formulation can affect ice crystallization by influencing the freezing point
and/or the crystallization mechanism. The sugars and milk salts
in ice cream mix affect the freezing point of the mix, which affects the degree of supercooling for a given coolant temperature
(Marshall and others 2003). Fat can reduce ice crystal size by
taking the place of water. In this case, for a given number of nuclei formed in ice cream that is hardened to a given temperature,
there will be less water to freeze when fat replaces some of the
water (Hartel 1996). Partial coalescence of fat globules may also
have a role in keeping ice crystals small (Barfod 2001). Proteins,
having a high molecular weight, diffuse more slowly than other
solutes, which could cause them to interfere with the lattice incorporation step of ice crystallization. The air cells may act as an
insulator, mitigating temperature fluctuations (Sofjan and Hartel
2004), and may physically impede ice crystal growth.
Stabilizers in ice cream attenuate the effects of recrystallization during storage, but have no significant effect in the freezer
itself (Goff 1997; Bolliger and others 2000). They neither change
the thermodynamics of the ice cream mix nor alter ice nucleation kinetics, though they do slow down crystal growth during
heat shock (Muhr and others 1986; Marshall and others 2003).
Emulsifiers help destabilize fat, incorporate more and smaller air
bubbles, and form thinner lamellae between air bubbles, all of
which are physical impediments to ice crystal growth (Marshall
and others 2003). In the SSF, emulsifiers have been shown to
lower the average ice crystal size (Hartel, unpublished).
Ice structuring proteins, ISPs (Regand and Goff 2006), and
propylene glycol monostearate (PGMS) (Aleong and others 2008)
are emerging as new options to limit ice crystal growth and recrystallization. Taken from freeze-tolerant plants, ISPs are believed to
act by adsorbing to the ice crystal surface and thus interfering

Vol. 9, 2010COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETY

219

CRFSFS: Comprehensive Reviews in Food Science and Food Safety

with the normal kinetics of water diffusion and adsorption at the
ice crystal-solution interface (Regand and Goff 2006). PGMS is
also thought to directly interfere with the lattice incorporation
step of crystallization and to retard ice crystal growth even more
strongly than ISPs (Aleong and others 2008).
Faster freezing results in smaller ice crystals. To that end, the
length of time that the ice cream spends in the SSF greatly impacts
the final ice crystal size distribution. A longer residence time
means that, for a given draw temperature, the ice cream mix is
slower to reach its exit temperature. It also gives the ice crystals in
the warmer bulk of the fluid more opportunity to recrystallize and
grow larger (Russell and others 1999; Drewett and Hartel 2007).
Additionally, a wide residence time distribution causes some of
the ice cream mix to spend a longer time in the freezer and
grow larger (Russell and others 1997). Decreased dasher speed,
fewer or smaller gaps between the dasher blades, and increased
shear thinning consistency in the fluid all are known to decrease
axial mixing within the freezer and to narrow the residence time
distribution. Ideally, no backmixing via plug flow would give the
narrowest distribution (Russell and others 1997).
To achieve a high cooling rate, heat transfer at the wall must be
as efficient as possible. Because heat travels more slowly through
ice than stainless steel, ice building up on the barrel wall acts
as an insulator and slows its own growth. Efficient scraping, by
keeping the blade sharp and close to the wall, ensures that heat
transfer remains fast, giving smaller ice crystals (Marshall and
others 2003).
The wall temperature has a direct effect on the cooling rate
of the mix: a higher driving force speeds initial nucleation and
growth. While wall temperature may not be the dominant factor
in determining the final ice crystal size distribution, lower wall
temperatures can lower the bulk temperature of the ice cream
faster, reducing residence time and improving the ice crystal size
distribution (Russell and others 1999; Drewett and Hartel 2007).
Dasher speed is known to have an effect on the final ice
crystal size; however, literature sources are not in agreement
as to this effect. Increasing dasher speed has been found to increase (Russell and others 1999), slightly increase (Drewett and
Hartel 2007), not to affect (Koxholt and others 2000), and to have
mixed effects (Inoue and others 2008) on ice crystal size. Crystal size could increase with dasher speed because of the extra
mechanical energy and frictional heat being put into the system
via the rotation of the dasher, which accelerates recrystallization
(Russell and others 1999). Conversely, increasing dasher speed
could help reduce ice crystal size by scraping off a thinner ice
layer. According to Schwartzbergs theory (Schwartzberg 1990), a
higher scraping rate should interrupt crystal growth and result in
smaller crystals being mixed into the bulk, but he also acknowledged that scraping produces frictional heat and that a dasher
must input some power into the freezer to agitate the product.
It is possible that dasher speed itself is not a direct predictor of
ice crystal size, instead the actual frictional heat generation may
give a better correlation with ice crystal size. This case suggests
an optimum balance between rotation rate and heat generation,
which would depend on the individual freezer and ice cream mix
composition to produce the smallest ice crystals.
Dashers are made in a variety of displacements, the percentage of the freezer barrel volume occupied by the dasher. Solid
dashers have high displacement, typically about 80%, whereas
open dashers have low displacement and often include beaters
(Marshall and others 2003). The effect of having a high displacement dasher is to bring all the ice cream mix closer to the wall
by making the mix pass through an annular space, which can
also reduce the residence time within the freezer and speed heat
transfer. High displacement, therefore, theoretically results in a
smaller ice crystal size (Marshall and others 2003). Using differ220

ent dasher displacements with a constant exit or wall temperature

has not been found, however, to have a significant effect on the
resulting ice crystal size distribution (Russell and others 1999).
Draw temperature is determined by the transport properties of
the system: heat transfer rates, extent of crystallization, frictional
heating, and so forth. A lower draw temperature results in less
time for hardening and, thus, smaller ice crystals. With conventional ice cream freezing, however, if the draw temperature is too
low, the ice cream will be too viscous to package into its retail
container (Marshall and others 2003). Also, a lower wall temperature lowers the bulk temperature inside the freezer, reducing
ripening (Koxholt and others 2000; Drewett and Hartel 2007).
A new technology for reducing draw temperature and eliminating the need for hardening is cold extrusion (Windhab and
Wildmoser 2002). The process reportedly improves both the
sensory and shelf-life properties of the ice cream. The lowtemperature extruder is placed after the normal scraped-surface
freezer and can bring the draw temperature down to 15 to
18 C (Windhab and Wildmoser 2002). The extruder may be in
either a single-screw configuration, which churns the ice cream
around the screw, or a twin-screw configuration, which kneads
and churns the ice cream between 2 parallel screws. Both operate at a rotational rate of about 15 rpm (Kunzig 2004). In the
extruder, high homogeneous local shear breaks up the fat globules, ice crystals, and air cells into smaller units. Because of the
smaller dispersion sizes, ice cream flowability is maintained. For
ice crystals, the low temperatures and high ice-cream viscosity
in the extruder not only help form additional secondary nuclei,
but the shear breaks apart ice aggregates, too. Most importantly,
fast heat transfer helps prevent local ice remelting of small crystals and recrystallization. Compared to conventional ice cream
freezing and hardening, including a twin-screw cold extruder reduces ice crystal and air cell size by a factor of 2-3 (Windhab
and Wildmoser 2002). The single-screw setup only had a small
effect on ice crystal size, but still impacted air cell size similar to
the twin-screw (Wildmoser 2004). The smaller fat globule aggregates resulting from the high shear in the cold extruder impart a
creamy texture to the ice cream without a buttery defect, making the technology useful for low-fat formulations (Windhab and
Wildmoser 2002).
One potential modification to the typical ice cream freezing
process is to recirculate a portion of the product stream back
to the freezer inlet stream. The benefit of this practice is mainly
for start-up, where recirculation minimizes production of unused
unfrozen product. The recirculation stream brings existing ice
crystals into the mix where they can serve as seed crystals and
can continue to grow in size (Hartel 1996). This results in a wider
residence time distribution with a higher average residence time
and, thus, a larger, wider crystal size distribution (Schwartzberg
1990; Hartel 1996; Levenspiel 1998; Kusumaatmaja 2009). Recirculation does have a beneficial effect on the air cell size, however, because the extra whippingprovided by the increased
residence timebreaks the air cells into smaller bubbles (Chang
and Hartel 2002; Goff 2002; Thakur and others 2005).
Preaeration, or prewhipping, is the practice of whipping the
ice cream mix to incorporate air and begin destabilizing the fat
before the mix is frozen (Marshall and others 2003). Preaeration
reportedly gives a slightly smaller air bubble size and improves
perceptions of creaminess and smoothness (Burmester and others
2005). Smaller air cells pack more tightly, which leaves smaller
spaces between bubbles for ice to grow in, and the ice crystals,
consequently, are smaller in size (Barfod 2001). Smaller air bubble size, in addition to small ice crystal size, can function as a
fat replacer (Tharp 1997). A recent study (Kusumaatmaja 2009),
however, found little correlation between pre-aeration and ice
crystal size or air cell size. Air cell size likely did not decrease

COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETYVol. 9, 2010

Mechanisms of ice crystallization in ice cream . . .

significantly with preaeration due to relatively low shear before References
the mix was frozen compared to the higher shear experienced Aleong JM, Frochot S, Goff HD. 2008. Ice recrystallization inhibition in ice cream by propyinside the freezer barrel. The fact that air cell size did not change lene glycol monostearate. J Food Sci 73(9):E4638.
NM. 2001. The emulsifier effect. Dairy Ind Int 66(1):323.
with preaeration was likely why ice crystal size did not change ei- Barfod
Ben-Yoseph E, Hartel RW. 1998. Computer simulation of ice recrystallization in ice cream
ther (Kusumaatmaja 2009). Also, without the ice crystals present during storage. J Food Eng 38(3):30929.
to provide viscosity and stability to the ice cream structure, air Berger KG, Bullimore BK, White GW, Wright WB. 1972. The structure of ice cream: part 1.
Dairy Ind 37(8):41925.
bubbles are prone to coalescence, reducing the impact of preaer- Bolliger S, Wildmoser H, Goff HD, Tharp BW. 2000. Relationships between ice cream mix
viscoelasticity and ice crystal growth in ice cream. Int Dairy J 10(11):7917.
ation (Windhab and Wildmoser 2002).
RL. 1984. Plotting freezing curves for frozen desserts. Dairy Record 85(7):8687.
An emerging technology in ice cream that has yet to be fully Bradley
Burmester SSH, Russell AB, Cebula DJ. 2005. The evolution of ice cream technology. New
developed is the use of ultrasonic waves to decrease ice crys- Food 2:4245.
tal size (Zheng and Sun 2006). High intensity ultrasound causes Caldwell KB, Goff HD, Stanley DW. 1992. A low-temperature scanning electron-microscopy
study of ice cream 1. Techniques and general microstructure. Food Struct 11(1):19.
cavitationnegative acoustic pressure causing the liquid to frac- Cebula
DJ, Russell AB. 1998. Ice crystallization control in ice cream. In: Buchheim W, editor.
ture and form bubblesand microstreamingvigorous circula- 1997 International Dairy Federation international ice cream symposium. Proceedings of
the
international
symposium; 1997 Sept 1819; Athens, Greece. Brussels, Belgium: Intl.
tion of the cavitation bubbles through the fluid causing eddies.
Dairy Federation. p 1319.
Cavitation can help increase nucleation and help prevent ice en- Chang
Y. 2000. Air cell measurement in dairy foam systems [Ph.D. thesis]. Madison, Wis.:
crustation on the wall, and microstreaming can improve heat and Univ. of Wisconsin - Madison. 296 p.
Chang
Y,
Hartel RW. 2002. Development of air cells in a batch ice cream freezer. J Food Eng
mass transfer. The acoustic stress also causes ice crystals to frac55(1):7178.
ture, creating smaller crystals. It was suggested that ultrasound Donhowe
DP, Hartel RW. 1996a. Recrystallization of ice during bulk storage of ice cream.
could be applied to ice cream production by attaching a trans- Int Dairy J 6(1112):120921.
Donhowe
DP, Hartel RW. 1996b. Recrystallization of ice in ice cream during controlled
ducer coupling to the external surface of the SSF barrel (Zheng
accelerated storage. Int Dairy J 6(1112):1191208.
and Sun 2006).
Drewett EM, Hartel RW. 2007. Ice crystallization in a scraped surface freezer. J Food Eng

Conclusions
The final structure and sensory properties of ice cream are a
function of many different processing and formulation variables
that tend to be a complex function of one another. Ice cream companies today already are able to produce high-quality ice cream.
However, with the ongoing quest to make products healthier,
longer-lasting, cheaper, and/or of higher quality, understanding
the effects of the numerous variables and their interactions helps
both scientists and engineers. This understanding allows scientists to better predict the effects of processing and formulation
changes and allows engineers to optimize plans and to make
improved assumptions in models.
Existing research has made great strides in elucidating the
mechanisms of ice crystallization in the ice cream freezer, though
more research is still needed to obtain the definitive model. The
vast majority of research has looked at ice crystallization in model
solutions in model systems. Ideally, scientists will one day be able
to piece together how ice forms in ice cream mix in its actual system.
Nomenclature
Roman letters
C = Concentration, g solute/g solution; C g , maximum
freeze concentration
n = Variable parameter, dimensionless
r = Radius, m; r, average radius; r0 , initial average
radius
R = Recrystallization rate, m/s1/n
t = Time, s
T = Temperature, K; T g , glass transition temperature;
T g , glass transition temperature at maximum
freeze concentration; T m , onset of melting at
maximum freeze concentration; T ,
temperature at infinite radius
Greek letters
H fus = Latent heat of crystallization, J/kg
T  = Freezing point depression, K
s = Crystal density, kg/m3
= Interfacial tension (surface free energy per unit
area), J/m2 or N/m
= Variable parameter, s/mn

78(3):10606.
Eisner MD, Wildmoser H, Windhab EJ. 2005. Air cell microstructuring in a high-viscous ice
cream matrix. Colloids Surf A 263(13):3909.
Fennema OR. 1973. Low-temperature preservation of foods and living matter. New York:
Marcel Dekker. 598 p.
Flemings MC. 2005. Coarsening in solidification processing. Mater Trans 46(5):895900.
Glicksman ME, Voorhees PW. 1984. Ostwald ripening and relaxation in dendritic structures.
Metall Trans 15A(5):9951001.
Goff HD. 1997. Colloidal aspects of ice cream: a review. Int Dairy J 7(67):36373.
Goff HD. 2002. Formation and stabilisation of structure in ice cream and related products.
Curr Opin Colloid Interface Sci 7(56):4327.
Goff HD, Hartel RW. 2006. Ice cream and frozen desserts. In: Hui YH, editor. Handbook of
food science, technology, and engineering, ch 154. Boca Raton, Fla.: CRC Press. p 148.
Hagiwara T, Hartel RW. 1996. Effect of sweetener, stabilizer, and storage temperature on ice
recrystallization in ice cream. J Dairy Sci 79(5):73544.
Hartel RW. 1996. Ice crystallization during the manufacture of ice cream. Trends Food Sci
Technol 7(10):31521.
Hartel RW. 1998a. Phase transitions in ice cream. In: Rao MA, Hartel RW, editors. Phase/state
transitions in foods: chemical, structural, and rheological changes. IFT basic symposium
series. New York: Marcel Dekker. p 32768.
Hartel RW. 1998b. Mechanisms and kinetics of recrystallization in ice cream. In: Reid DS,
editor. The properties of water in foods: ISOPOW 6. London, U.K.: Blackie Academic and
Professional. p 287319.
Hartel RW. 2001. Crystallization in foods. Gaithersburg, Md.: Aspen Publishers. 325 p.
Hindmarsh JP, Russell AB, Chen XD. 2005. Measuring dendritic growth in undercooled
sucrose solution droplets. J Cryst Growth 285(12):23648.
Hobbs PV. 1974. Ice physics. Oxford, U.K.: Clarendon Press. 837 p.
Huang SC, Glicksman ME. 1981a. Fundamentals of dendritic solidificationII. Development
of sidebranch structure. Acta Metall 29:71734.
Huang SC, Glicksman ME. 1981b. Fundamentals of dendritic solidificationI. steady-state tip
growth. Acta Metall 29:70115.
Inoue K, Ochi H, Taketsuka M, Saito H, Sakurai K, Ichihashi N, Iwatsuki K, Kokubo S. 2008.
Modeling of the effect of freezer conditions on the principal constituent parameters of ice
cream by using response surface methodology. J Dairy Sci 91(5):172232.
Jacobsen JA. 1999. Ice crystallization in a scraped-surface slush freezer [MSc thesis]. Madison,
Wis.: University of Wisconsin - Madison. 92 p.
Jain SC, Hughes AE. 1978. Ostwald ripening and its application to precipitates and colloids
in ionic crystals and glasses. J Mater Sci 13(8):161131.
Kahlweit M. 1975. Ostwald ripening of precipitates. Adv Colloid Interface Sci 5(1):135.
Koxholt M, Eisenmann B, Hinrichs J. 2000. Effect of process parameters on the structure of
ice cream. Eur Dairy Mag 1:2730.
Kunzig R. 2004. Less fat, better taste. Discover 25(6):267.
Kusumaatmaja W. 2009. Effects of mix pre-aeration and product recirculation on ice cream
microstructure and sensory qualities [MSc thesis]. Madison, WI: University of Wisconsin Madison. 136 p.
Levenspiel O. 1998. Chemical reaction engineering. 3rd ed. Hoboken, N.J.: Wiley. 688 p.
Lian G, Moore S, Heeney L. 2006. Population balance and computational fluid dynamics
modelling of ice crystallisation in a scraped-surface freezer. Chem Eng Sci 61(23):781926.
Libbrecht KG. 2005. The physics of snow crystals. Rep Prog Phys 68(4):85595.
Lifshitz IM, Slyozov VV. 1961. The kinetics of precipitation from supersaturated solid solutions. J Phys Chem Solids 19(12):3550.
Macklin WC, Ryan BF. 1966. Habits of ice grown in supercooled water and aqueous solutions. Philos Mag 14(130):84760.
Marshall RT, Goff HD, Hartel RW. 2003. Ice cream. 6th ed. New York: Kluwer Academic/Plenum Publishers. 371 p.
McCabe WL, Smith JC, Harriott P. 2005. Unit operations of chemical engineering. 7th ed.
Boston, Mass.: McGraw-Hill. 1140 p.
Muhr AH, Blanshard JMV, Sheard SJ. 1986. Effects of polysaccharide stabilizers on the
nucleation of ice. J Food Technol 21(5):587603.
Mullin JW. 1993. Crystallization. 3rd ed. Oxford, U.K.: Butterworth-Heineman. 527 p.

Vol. 9, 2010COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETY

221

CRFSFS: Comprehensive Reviews in Food Science and Food Safety

Mullis AM. 2006. An analytical geometrical model for secondary dendrite arm detachment.
Scr Mater 54(5):7959.
Omran AM, King CJ. 1974. Kinetics of ice crystallization in sugar solutions and fruit juices.
AIChE J 20(4):795803.
Regand A, Goff HD. 2006. Ice recrystallization inhibition in ice cream as affected by ice
structuring proteins from winter wheat grass. J Dairy Sci 89(1):4957.
Russell AB, Burmester SSH, Winch PJ. 1997. Characterization of shear thinning flow within
a scraped-surface heat exchanger. Food Bioprod Process 75(C3):1917.
Russell AB, Cheney PE, Wantling SD. 1999. Influence of freezing conditions on ice crystallisation in ice cream. J Food Eng 39(2):17991.
Schawe JEK. 2006. A quantitative DSC analysis of the metastable phase behavior of the
sucrose-water system. Thermochim Acta 451(12):11525.
Schwartzberg HG. 1990. Food freeze concentration. In: Rao MA, Schwartzberg HG, editors. Biotechnology and food process engineering. IFT basic symposium series. New York:
Marcel Dekker. p 127201.
Schwartzberg HG, Liu Y. 1990. Ice crystal growth on chilled scraped surfaces. In: Am Inst
Chem Eng Summer National Meeting; 1990 Aug 1922, San Diego, CA. Paper nr 2g.
42 p.
Si W. 2000. Mechanisms of ice crystallization in scraped-surface heat exchanger [MSc thesis].
Madison, Wis.: University of Wisconsin - Madison. 98 p.
Smith CE, Schwartzberg HG. 1985. Ice crystal size changes during ripening in freeze concentration. Biotechnol Progr 1(2):11120.
Sodawala S, Garside J. 1997. Ice nucleation on a cold surface. In: The 1997 Jubilee Research
Event; 1997 Apr 89; Nottingham, UK. Rugby, U.K.: Institution of Chemical Engineers.
p 47780.

222

Sofjan RP, Hartel RW. 2004. Effects of overrun on structural and physical characteristics of
ice cream. Int Dairy J 14(3):25562.
Stocking JH, King CJ. 1976. Secondary nucleation of ice in sugar solutions and fruit juices.
AIChE J 22(1):13140.
Sutton RL, Evans ID, Crilly JF. 1994. Modeling ice crystal coarsening in concentrated disperse
food systems. J Food Sci 59(6):122733.
Teraoka Y, Saito A, Okawa S. 2002. Ice crystal growth in supercooled solution. Int J Refrig
25(2):21825.
Thakur RK, Vial C, Djelveh G. 2005. Combined effects of process parameters and composition
on foaming of dairy emulsions at low temperature in an agitated column. J Food Eng
68(3):33547.
Tharp BW. 1997. On to the 21st century: possibilities of new technologies for the ice cream
industry. Eur Dairy Mag 1:610.

Wagner C. 1961. Theorie der Alterung von Niederschlagen durch Umlosen.

Z Elektrochemie
65(78):58191.
Wildmoser J. 2004. Impact of low-temperature extrusion processing on disperse microstruc
ture in ice cream systems [DTechSci]. Zurich,
Switzerland: Swiss Federal Inst. of Technology,

Zurich.
132 p.
Windhab EJ, Wildmoser H. 2002. Extrusion: a novel technology for the manufacture of ice
cream. Bull Int Dairy Fed 374:439.
Zheng L. 2006. Heat transfer during freezing on a scraped surface [Ph.D. thesis]. Auckland,
New Zealand: University of Auckland. 179 p.
Zheng L, Sun DW. 2006. Innovative applications of power ultrasound during food freezing
processes: a review. Trends Food Sci Technol 17(1):1623.

COMPREHENSIVE REVIEWS IN FOOD SCIENCE AND FOOD SAFETYVol. 9, 2010